A first test-print protocol

Before you print a real chip, print a coupon. It takes about an hour including washing and curing, and it replaces three guesses with three measurements:

your real minimum channel size;
your real minimum wall between channels;
whether your ports fit the fittings you own.

Every manufacturability warning OpenChip shows you is measured against numbers you typed in. This is how you make them true.

The coupon

Build it in OpenChip as a normal project. About 40 × 60 mm, two-piece, so it is quick to print and trivial to drain.

Part 1 — a channel ladder

Eight straight channels side by side, each with a port at both ends, widths stepping from 0.10 mm to 0.60 mm in 0.05 mm steps (drop the finest two if you are on a coarse machine).

Space them 3 mm apart so this part tests width only.

Use a dimension marker at each end of each channel set to the same width, so each channel is uniform.

Expect channel-below-min-feature warnings on the narrow ones. That is the point — you are testing whether the current profile is right.

Part 2 — a wall ladder

Six pairs of 0.5 mm channels, each pair with the gap between them stepping from 0.2 mm to 0.7 mm in 0.1 mm steps.

Each channel needs its own two ports so you can flush them independently — this is what lets you find out whether a pair has merged.

Expect channels-too-close on the tight pairs.

Part 3 — ports

One of each fitting you own, on its own short channel: a luer slip, a luer lock, a tubing socket.

Give each a partner port so you can flush it.

Print it

Flat on the plate, channel side up. No supports. Your normal exposure settings — the point is to characterise the process you actually use.

Print the lid too, but do not bond it yet.

Wash it

Exactly as you would a real chip. Drain, flush, wash, flush, dry. See Washing and post-curing.

Do not post-cure yet — inspect first.

Measure it

Minimum channel size

Look down each channel in the ladder against a strong light. Then, with the lid still off, run a fine wire or a drop of dyed water along each groove.

The narrowest channel that is open along its whole length is your real minimum feature size.

Expect it to be larger than the rule of thumb suggested. If the rule said 0.2 mm and you measure 0.25 mm, use 0.25 mm.

Also measure the actual width of a couple of channels with calipers or a microscope, against what you drew. The difference is your dimensional offset, and it is repeatable — see Resolution vs. channel size.

Minimum wall

Bond the lid on the wall-ladder section — or just tape a piece of acrylic over it temporarily — and flush one channel of each pair with dyed water while watching the other.

Dye appearing in the neighbour means the wall cured through and the two merged.

The narrowest gap where no dye crosses is your real minimum printable wall. The default is 0.4 mm; a clear resin will do worse and a well-characterised pigmented one may do better.

Ports

Try each fitting.

Too tight — the fitting will not seat. Open the socket by 0.05 mm and reprint that part.
Too loose — it leaks or falls off. Reduce by 0.05 mm.
Right — it seats firmly with a slight push and does not leak under a syringe.

Note the offset. It stays good until you change resin.

Put the numbers in

Settings → Printer, and set minFeatureSizeMm to what you measured.

Design checks → Thresholds, and set Minimum printable wall to what you measured.

The port offset is not a setting — apply it to the port dimensions when you draw, or save a project with corrected ports as your own starting template.

Then post-cure the coupon

Keep it. It is a reference for the next time you change resin, and a quick way to tell whether a new bottle behaves like the old one.

Repeat when

You change resin — the most important one.
You change machine.
You replace the FEP, if you are working near the limit.
Prints start failing in a way they did not before.

Why this is worth an hour

Every warning OpenChip gives you is arithmetic against numbers you supplied. With guessed numbers you get warnings that are either too cautious — so you ignore them — or too permissive, so they let a chip through that will not print.

With measured numbers the panel becomes genuinely predictive, and that is worth more than any single chip.